CN116465627A

CN116465627A - Mechanical equipment fault monitoring system

Info

Publication number: CN116465627A
Application number: CN202310359772.3A
Authority: CN
Inventors: 朱冠华; 张清华; 孙国玺; 蔡业彬; 胡绍林; 荆晓远; 张磊
Original assignee: Guangdong University of Petrochemical Technology
Current assignee: Guangdong University of Petrochemical Technology
Priority date: 2023-04-04
Filing date: 2023-04-04
Publication date: 2023-07-21

Abstract

The invention provides a mechanical equipment fault monitoring system, comprising: the data acquisition end acquires vibration data of the mechanical equipment and sends the vibration data to the data processing end, and the data processing end divides the first dimensionality parameter of the first period of the vibration data by the second dimensionality parameter of the second period to obtain an abnormal period dimensionless index so as to generate early warning information when the abnormal period dimensionless index is larger than a preset early warning threshold value; the starting time of the second period is earlier than the starting time of the first period, the ending time of the second period is earlier than or equal to the ending time of the first period, and the time length of the second period is longer than or equal to the time length of the first period. The invention can obtain the dimensionless index of the different periods by selecting different dimensionless parameters of the different periods to be used for analyzing the faults of the mechanical equipment, effectively inhibit the interference of external environment and improve the defect of untimely early warning.

Description

Mechanical equipment fault monitoring system

Technical Field

The invention relates to the field of mechanical equipment, in particular to a mechanical equipment fault monitoring system.

Background

Industrial units are complex multivariable systems, rotary machines are one of the core equipment of the industrial units, the mechanical equipment bears the development of most heavy industries in China, the mechanical equipment is assembled by various different parts, the mechanical equipment usually works under the conditions of high load and multiple working conditions, the complex working conditions greatly increase the potential safety hazards of the equipment, if some elements or component equipment are abnormal or have faults, the abnormal operation working conditions of the equipment can be caused, and the equipment can be stopped for maintenance or even safety accidents can be caused.

The fault diagnosis method commonly used by the rotating machinery at present is to mine out the fault characteristics of equipment, such as dimensional time domain characteristics, through an effective signal processing technology. However, in the actual monitoring process, the dimensional time domain features are easy to be disturbed, when the unit is disturbed at a certain moment, especially when the unit is disturbed in a certain time period, the time domain features measured in the current time period have obvious change in a very large probability, and when monitoring or diagnosing is performed by using a method based on the time domain features, the system can be wrongly warned, and unnecessary influence is generated. In addition, due to certain coupling among units forming the equipment, the rotary machine can generate serious faults in a short time after early warning based on the conventional technology, emergency treatment time cannot be reserved for staff, and the defect of untimely early warning exists.

Disclosure of Invention

The embodiment of the invention provides a mechanical equipment fault monitoring system, which can obtain different-period dimensionless indexes by selecting different dimensional parameters of different periods to be used for analyzing mechanical equipment faults, effectively inhibit interference of external environments and improve the defect of untimely early warning.

In order to achieve the above object, an embodiment of the present invention provides a mechanical equipment fault monitoring system, including:

the data acquisition end is used for acquiring vibration data of the mechanical equipment;

the data processing end is used for acquiring the vibration data from the data acquisition end, dividing the first dimensionality parameter of the first period of the vibration data by the second dimensionality parameter of the corresponding second period to obtain an abnormal period dimensionless index; wherein the start time of the second period is earlier than the start time of the first period, the end time of the second period is earlier than or equal to the end time of the first period, and the time length of the second period is longer than or equal to the time length of the first period;

the data processing end is further used for generating early warning information when the abnormal period dimensionless number index is larger than a preset early warning threshold value.

As an improvement of the above scheme, the aperiodic dimensionless index is an aperiodic peak value index, the first dimensionality parameter is a signal peak value, and the second dimensionality parameter is a root mean square value; the off-period peak indicator is equal to a signal peak of the vibration data at the first period divided by a root mean square value of the vibration data at a corresponding second period.

As an improvement of the above scheme, the aperiodic dimensionless index is an aperiodic pulse index, the first dimensionality parameter is a signal peak value, and the second dimensionality parameter is an absolute average value; the off-period pulse index is equal to the signal peak value of the vibration data in the first period divided by the absolute average value of the vibration data in the corresponding second period.

As an improvement of the scheme, the aperiodic dimensionless index is an aperiodic margin index, the first dimensionality parameter is a signal peak value, and the second dimensionality parameter is a mean square value; the abnormal period margin index is equal to a signal peak value of the vibration data in the first period divided by a mean square value of the vibration data in a corresponding second period.

As an improvement of the above scheme, the aperiodic dimensionless index is an aperiodic waveform index, the first dimensionality parameter is a root mean square value, and the second dimensionality parameter is an absolute average value; the abnormal-period waveform index is equal to a root mean square value of the vibration data in the first period divided by an absolute average value of the vibration data in a corresponding second period.

As an improvement of the above scheme, the abnormal period dimensionless index is an abnormal period kurtosis value ratio index, and the abnormal period waveform index is equal to an average value of kurtosis values of the vibration signal in the first period divided by an average value of kurtosis values of the corresponding second period.

As an improvement of the above solution, the data processing end is further configured to obtain the first period, the second period, and the early warning threshold by:

constructing an original abnormal period monitoring model;

the method comprises the steps of carrying out model training on the different-period monitoring model according to a preset particle swarm algorithm and a pre-acquired historical vibration data sample to obtain a first period, a second period, a period interval and an early warning threshold;

the position parameter of the particle swarm algorithm is the first period, the second period and the period interval, the adaptability is early warning time of particles, and the period interval is a time interval between the starting time of the first period and the ending time of the corresponding second period.

As an improvement of the above scheme, the data processing end is further configured to apply the abnormal period monitoring model to online monitoring of the mechanical device, and update the model periodically according to an online monitoring result.

As an improvement of the above solution, the data processing end is further configured to:

if the online monitoring does not generate early warning information and the early warning threshold value is larger than a first preset threshold value within a preset time period after the last model updating, reducing the early warning threshold value by a first preset step length;

And if the online monitoring is in a false alarm phenomenon and the early warning threshold is smaller than a second preset threshold within a preset time period after the last model updating, increasing the early warning threshold by a second preset step length.

As an improvement of the scheme, the data acquisition end comprises a speed sensor, a signal conditioning module, a signal filtering module and an acquisition card;

the two probes of the speed sensor are respectively arranged in the horizontal direction and the vertical direction of the rotary mechanical bearing cover and are used for converting physical quantity generated by rotary mechanical vibration into an electric signal and transmitting the electric signal to the signal conditioning module;

the signal conditioning module is used for performing a method on the electric signal and outputting the amplified electric signal to the signal filtering module;

the signal filtering module is used for performing high-frequency signal filtering operation on the electric signals and transmitting the filtered electric signals to the acquisition card;

the acquisition card acquires and processes the electric signals and uploads the electric signals to the data processing end by utilizing a TCP/IP protocol.

As an improvement of the scheme, the data acquisition end is a triaxial acceleration sensor or a wireless vibration sensor.

Compared with the prior art, the mechanical equipment fault monitoring system provided by the embodiment of the invention obtains the abnormal period dimensionless index by obtaining the vibration data of the mechanical equipment and dividing the first dimensionality parameter of the first period of the vibration data by the second dimensionality parameter of the second period; the starting time of the second period is earlier than the starting time of the first period, the ending time of the second period is earlier than or equal to the ending time of the first period, and the time length of the second period is longer than or equal to the time length of the first period. According to the embodiment of the invention, the different dimensionless indexes of different periods are obtained by selecting different dimensionless parameters of different periods to be used for analyzing the faults of mechanical equipment, so that the interference of external environment is effectively inhibited, and the defect of untimely early warning is improved.

Drawings

FIG. 1 is a flow chart of a method of a system for monitoring a fault in a machine according to an embodiment of the present invention;

FIG. 2 is a schematic diagram of period division according to an embodiment of the present invention;

FIG. 3 is a schematic diagram of period division according to an embodiment of the present invention;

FIG. 4 is a schematic diagram of period division according to an embodiment of the present invention;

FIG. 5 is a schematic diagram of a rotary machine fault data model provided in an embodiment of the present invention;

FIG. 6 is a graph of an aperiodic peak index and a fault data model provided by one embodiment of the present invention;

FIG. 7 is a schematic diagram of period division according to an embodiment of the present invention;

FIG. 8 is a diagram of raw data distribution provided by an embodiment of the present invention;

FIG. 9 is a graph showing an aperiodic peak index data distribution according to one embodiment of the present invention;

FIG. 10 is a schematic diagram of rotor rub fault 1 displacement peak-to-peak according to an embodiment of the present invention;

FIG. 11 is a schematic diagram of rotor rub fault 2 shift peak-to-peak according to an embodiment of the present invention;

FIG. 12 is a schematic diagram of rotor rub fault 3 shift peak-to-peak according to an embodiment of the present invention;

FIG. 13 is a schematic diagram of misalignment fault 1 shift peak-to-peak provided by an embodiment of the present invention;

FIG. 14 is a schematic diagram of misalignment fault 2 shift peak-to-peak provided by an embodiment of the present invention;

FIG. 15 is a schematic diagram of misalignment fault 3 shift peak-to-peak provided by an embodiment of the present invention;

FIG. 16 is a schematic diagram of peak-to-peak displacement of station 1 according to an embodiment of the present invention;

FIG. 17 is a schematic diagram of peak-to-peak displacement at station 2 according to one embodiment of the present invention;

FIG. 18 is a graph showing peak-to-peak displacement at station 3 according to one embodiment of the present invention;

FIG. 19 is a schematic diagram of peak-to-peak displacement at station 4 according to one embodiment of the present invention;

FIG. 20 is a schematic diagram of peak-to-peak displacement of a measurement point 5 according to an embodiment of the present invention;

FIG. 21 is a schematic diagram of peak-to-peak displacement of station 6 according to an embodiment of the present invention;

FIG. 22 is a schematic diagram of peak-to-peak displacement of station 7 according to an embodiment of the present invention;

FIG. 23 is a schematic diagram of peak-to-peak displacement at station 8 according to one embodiment of the present invention;

FIG. 24 is a boundary view of a turbine and generator evaluation area provided by an embodiment of the present invention;

FIG. 25 is a schematic diagram of a displacement peak-to-peak of a training set 140529044807131 provided by an embodiment of the invention;

FIG. 26 is a schematic diagram of a displacement peak-to-peak of a test set group 170318044053924 provided by an embodiment of the invention;

FIG. 27 is raw data of a misalignment fault provided by an embodiment of the present invention;

FIG. 28 is a machine peak-to-peak schematic diagram of a misalignment fault provided by an embodiment of the present invention;

FIG. 29 is a schematic diagram of a system architecture according to an embodiment of the present invention;

FIG. 30 is a schematic diagram of a system architecture according to an embodiment of the present invention;

fig. 31 is a schematic diagram of a system structure according to an embodiment of the present invention.

Detailed Description

The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.

The embodiment of the invention provides a mechanical equipment fault monitoring system, which comprises:

Referring to fig. 1, a method flow chart of a fault monitoring system for a mechanical device according to an embodiment of the present invention is provided. The method comprises the steps of S11 to S13:

s11, a data acquisition end acquires vibration data of mechanical equipment;

s12, the data processing end acquires the vibration data from the data acquisition end, and divides the first dimensionality parameter of the first period of the vibration data by the second dimensionality parameter of the corresponding second period to obtain a non-dimensionality index of the different period; wherein the start time of the second period is earlier than the start time of the first period, the end time of the second period is earlier than or equal to the end time of the first period, and the time length of the second period is longer than or equal to the time length of the first period;

and S13, when the abnormal period dimensionless number index is larger than a preset early warning threshold value, the data processing end generates early warning information.

For example, referring to the original vibration displacement data curve shown in fig. 2, the abscissa is time, the ordinate is original vibration displacement data, the original vibration displacement data (vibration data) of the mechanical device is obtained in real time, and the period division is performed as shown in fig. 2, t1 to t0 are the first periods, and t2 to t1 are the second periods corresponding to the first periods. When the dimensionless index of the different period is calculated in real time, the current time is taken as the ending time of the first period, the time is shifted forward (t 0-t 1) to obtain the starting time of the first period, the starting time of the first period is taken as the ending time of the second period, the time is shifted (t 1-t 2) to obtain the starting time of the second period, the first dimensionality parameter of the first period and the second dimensionality parameter of the second period are determined according to the data shown in fig. 2, then the first dimensionality parameter and the second dimensionality parameter of the first period are divided to obtain the dimensionless index of the different period, the first period is a short period, the second period is a long period, the molecules of the dimensionless index calculation formula of the different period use short period data to reflect the change and the change trend of the monitoring data in a short period, the denominator uses the long period data to reflect the change and the change trend of the relevant characteristics of the monitoring data in a long period, and the problem that the change of the data is not early warning in time can be solved by adopting the monitoring mode, and the monitoring effect of the monitoring mode can be enhanced to some extent.

Further, the end time of the second period is prior to or equal to the start time of the first period, or the start time of the first period is prior to the end time of the second period.

Specifically, the ending time of the second period is earlier than or equal to the starting time of the corresponding first period, the interval duration between the first period and the corresponding second period is a preset interval duration, the preset interval duration is greater than or equal to zero, the interval duration of the first period and the corresponding second period shown in fig. 2 is 0, the first period x1 is t1 'to t0' and the corresponding second period x3 is t3 'to t2', the preset interval duration x2 is (t 2'-t 1'), in this case, the data used by the numerator and the denominator of the calculation formula of the non-dimensional parameter of the different period are not overlapped, and interference caused by overlapping of the calculated data used by the numerator and the denominator is avoided. Alternatively, the start time of the first period is earlier than the end time of the corresponding second period, as shown in fig. 4, T1 is the first period, T2 is the corresponding second period, and in the case that T2 includes T1, the difference between T2 and T1 should be greater than the preset value.

Specifically, the aperiodic dimensionless index is an aperiodic peak value index, an aperiodic pulse index, an aperiodic margin index, an aperiodic waveform index or an aperiodic kurtosis value ratio index.

In one embodiment, the aperiodic dimensionless index is an aperiodic peak index, the first dimensionality parameter is a signal peak, and the second dimensionality parameter is a root mean square value; the off-period peak indicator is equal to a signal peak of the vibration data at the first period divided by a root mean square value of the vibration data at a corresponding second period.

In one embodiment, the aperiodic dimensionless number indicator is an aperiodic pulse indicator, the first dimensionality parameter is a signal peak, and the second dimensionality parameter is an absolute average; the off-period pulse index is equal to the signal peak value of the vibration data in the first period divided by the absolute average value of the vibration data in the corresponding second period.

In one embodiment, the aperiodic dimensionless number indicator is an aperiodic margin number indicator, the first dimensionality parameter is a signal peak value, and the second dimensionality parameter is a mean square value; the abnormal period margin index is equal to a signal peak value of the vibration data in the first period divided by a mean square value of the vibration data in a corresponding second period.

In one embodiment, the aperiodic dimensionless number indicator is an aperiodic waveform indicator, the first dimensionality parameter is a root mean square value, and the second dimensionality parameter is an absolute average value; the abnormal-period waveform index is equal to a root mean square value of the vibration data in the first period divided by an absolute average value of the vibration data in a corresponding second period.

In one embodiment, the off-period dimensionless index is an off-period kurtosis value ratio index, the off-period waveform index being equal to an average of kurtosis values of the vibration signal at the first period divided by an average of kurtosis values of the corresponding second period.

For example, when the aperiodic dimensionless index is an aperiodic peak index, in conjunction with fig. 3, t0' is the current time, t1', t2', t3' is the time before t0', and t0' > t1' > t2' > t3', the aperiodic peak index g (t) is defined to consist of the ratio of the maximum value of f (x) in the x1 period to the root mean square value of f (x) in the x3 period.

The numerator of the different period peak index of the present embodiment represents the maximum value in the short period x1 before the time t0', and the denominator represents the root mean square in the larger long period x3 before the short period x1 used for the numerator calculation. The short period x1 used for the molecular calculation is separated from the long period x3 used for the denominator calculation by a period x2, and in order to avoid interference caused by superposition of calculated data used for the molecular denominator, the period x2 is generally required to be more than or equal to 1. The different period peak value index can comprehensively consider the monitoring data characteristics of the long period running state and the short period change trend, not only reflects the current change in a short time, but also sensitively reflects the relative change of the change trend in the short time and the change trend of the whole data trend.

Illustratively, explaining fault analysis with a specific example to obtain original data, according to analysis of the original data of the rotary machine, basically maintaining two sections of change of the original data, wherein the first section is a fault formation stage (also called a micro fault stage), and absolute displacement starts to rise and then rises to a certain amplitude to maintain the micro change for a period of time; the second stage is a serious fault stage, the absolute displacement amplitude of the fault continuously rises on the basis of the original state, the absolute displacement amplitude rises to the stage that the machine has to be stopped, at the moment, the machine possibly breaks down and has to be stopped, and the absolute displacement starting amplitude of the first stage is set to be 10 according to analysis of the original monitoring data distribution of the rotating machine; normal operation during time period [0, 1000 ]; faults begin to form during time periods [1000, 2600 ]; the device hold amplitude 30 continues to operate in a light fault condition for a period of time [2600, 3600 ]; the equipment failure continues to become severe to the point that shutdown is necessary during the time period [3600, 5200 ]; thereby generating a rotary machine failure data model for training (see fig. 5 for data model):

setting the abnormal period data x1=2, x2=20 and x3=200 for the established fault mathematical model, calculating the abnormal period peak value index, referring to fig. 6, comparing the calculated abnormal period peak value index with the curve of the fault data model (original data), and displaying larger peaks in intervals [1000, 2600] formed by the beginning of unit faults, wherein the abnormal period dimensionless data are shown in the intervals [1000, 2600 ]; in the interval with larger faults [3600, 5200], the abnormal cycle dimensionless data shows smaller wave peaks; and shows the same flat form as the original data in the flat running phase. According to the comparison between the peak index of different period and the original data shown in fig. 6, the original data has severe amplitude change in [1000, 2600], and timely early warning is needed at this stage, and early warning can be performed when the peak index value of different period exceeds 1.5, namely the early warning threshold is 1.5.

When the aperiodic dimensionless index is an aperiodic pulse index, referring to fig. 2, an aperiodic pulse index is defined, and the formula is as follows:

wherein: t is t ₀ -t ₁ <t ₁ -t ₂ ，f(t ₀ ,t ₁ ) Is in the original data at (t ₀ ,t ₁ ) Amplitude of vibration in a period of time, max|f (t ₀ ,t ₁ ) I is (t) ₀ ,t ₁ ) The maximum value in the samples collected during the time period, the denominator, represents the absolute average value. When the data is processed by the different period method, the numerator and the denominator respectively select the data at different moments, the number of samples selected by the numerator is a small period, and the number of samples selected by the denominator is a large period. As shown in fig. 2, t ₀ The moment is the current moment, t ₀ ～t ₁ The length of the small period chosen for the molecule, i.e. the choice of t for the molecule ₀ ～t ₁ Carrying out maximum value calculation on the samples of the sample; t is t ₁ ～t ₂ The large period length selected for the denominator, i.e. t is selected for the denominator ₁ ～t ₂ An absolute average value is calculated for the samples of (a).

When the abnormal period dimensionless index is an abnormal period margin index, the present embodiment uses a mode of selecting a small range abnormal period ratio, that is, selecting a large period and a small period in a small range, combining fig. 4, taking the current time as the reference, selecting T1 as a small period before the current time, and selecting T2 as a large period. And substituting the small period T1 data into the numerator of the margin index formula, and substituting the large period T2 data into the denominator of the margin index formula to obtain the different period margin index:

When the different-period dimensionless index is a different-period waveform index, the waveform index is subjected to different periodicity, and a new dimensionless index is respectively built by the molecular denominator in different periods to serve as a new monitoring value, so that the interference on the system analysis process can be reduced. Referring to fig. 3, in which the iso-periodicity, small period x1=t0 '-t1' is a period starting at the current time t0 'and proceeding to the end of t1', the mean value of root mean square is calculated; the large period x3=t2 '-t3' is a period starting from time t2 'and ending from time t3' onwards, and an average value of the absolute average values is calculated; the offset is t1'-t2' =fixed value so that the size periods do not overlap. The root mean square, absolute average, and aperiodic waveform index were calculated by the following formula:

the root mean square value is given by:

the absolute mean formula is:

the abnormal period dimensionless waveform index formula is as follows:

when the aperiodic non-dimensional index is an aperiodic kurtosis value ratio index, a new non-dimensional index (an aperiodic kurtosis value ratio H) is constructed by performing offset, aperiodic and ratiometric processing on the traditional kurtosis value F, and the calculation formula is shown in the following in combination with FIG. 7:

wherein, H is the ratio index of the isocycle kurtosis value; f (n) is time-series data of the nth data point; f (x 1 ') is an average value of the kurtosis values of x1' (x1=t0 "-t1") cycles calculated backward starting from the current time t0 "; f (x 3 ') is that the current time t0' is taken as a starting point, the current time t0' is shifted backwards by t0' -t2' periods, the average value of the period kurtosis value of x3' is calculated, and the value of x ' satisfies the following relations of x1' < x2', x1' < x3'.

In one embodiment, the data processing end is further configured to obtain the first period, the second period, and the early warning threshold by:

constructing an original abnormal period monitoring model;

Specifically, training according to a pre-acquired historical vibration data sample to obtain an abnormal period monitoring model; the model parameters comprise a first period (small period), a second period (large period), a relation (period interval) between the first period and the second period and the early warning threshold value. The advantages of model training and heteroperiodic analysis are described below in five specific examples:

1. off-cycle monitoring for off-cycle peak indicator

The data adopted by the invention are provided by Shenyang blower group measurement and control technology limited company, and are the extracted 4 original data sets, namely rotor imbalance faults and shafting misalignment faults, with reference to the following table and the original data distribution diagram shown in fig. 8. The sampled data type is real-time vibration displacement data of the unit, the sampling interval is about 5min, the sampling time is 1s, and the sampling frequency is 1024Hz.

According to the national standard 'measurement and evaluation of radial vibration of rotating machinery rotating shaft' (GB/T11348.1-1999) for evaluating the relative radial vibration of rotating shaft, the evaluation area of peak-to-peak vibration magnitude can be divided into four evaluation areas for evaluation, which are respectively: region a: vibrations of newly delivered machines typically belong to this area. Region B: it is generally believed that machines with vibration amplitudes in this region may operate indefinitely for extended periods of time. Region C: machines having vibration amplitudes in this region are generally considered unsuitable for long-term continuous operation, and generally the machine can be operated in this state for a limited period of time until there is a suitable opportunity to take remedial action. Region D: vibration amplitude in this region is generally considered to be severe enough to cause machine damage. For long-running machines, it is necessary to set operating vibration limits for the machine, which take the form of alarms and stops. According to the national standard "measurement and assessment of radial vibrations of the rotating machine shaft" (GB/T11348.1-1999), the warning value should be higher than the baseline by a value equal to a fraction of the upper limit of zone B, and if the baseline is low, the warning value may be lower than zone C.

According to national standard early warning value setting standard, carrying out early warning value setting on each machine, firstly extracting peak-to-peak value characteristics of each data set, and comparing the peak-to-peak value characteristics with the early warning value to carry out equipment fault monitoring, wherein specific peak-to-peak value early warning value and early warning time are as follows:

Machine for processing a sheet of material	Early warning threshold	Early warning time
			Misalignment 1	432	6890
Misalignment 2	75	2580
			Misalignment 3	159	3035
Imbalance 1	50	3628

The PSO algorithm (particle swarm optimization) can be used for effectively searching the optimal heteroperiodic variables x1, x2 and x3 (see FIG. 3), and the running time for searching the optimal solution is greatly saved.

The PSO algorithm is used for searching the optimal variables x1, x2 and x3, and the main steps are as follows:

step 1: initial parameter determination

Firstly, setting the particle population to be 50; the search dimension is 3-dimensional, and the position parameter of each particle is (x 1, x2, x 3); the value range of x1 is [1,20], the value range of x2 is [1,50], and the value range of x3 is [1,100]; the limiting range of the speed is [ -1,1]; the initial position of the particles is (1,10,50); the iteration number of the algorithm is 500; the fitness is the early warning time of the particles in each device.

Step 2: calculating particle fitness

(1) Dimensionless treatment

The experiment uses the different period variables x1, x2 and x3 in the particles to extract different period peak index data of 4 original data sets.

(2) Dimensionless data mean filtering

Because of the noise interference in the original data, if the pre-warning value extraction is directly carried out through the dimensionless data, certain large noise data can be recognized as the pre-warning data, the pre-warning value extraction is greatly interfered, and the dimensionless data can be noise-reduced through mean value filtering. The mean value filtering is to establish a window with a certain length on the original time sequence, and take the mean value of the data in the window as the new value of the data at the end of the window.

(3) Calculating fitness

And extracting the time of the first peak from the dimensionless data after the mean value filtering as early warning time, namely, the fitness of a PSO algorithm.

Step 3: updating individual optimal fitness and population optimal fitness

And using the lowest fitness in the individual iteration history as the optimal fitness of the individual, and using the lowest fitness in the optimal fitness of all the particle individuals in the particle population as the optimal fitness of the population.

Step 4: particle position and velocity update

And updating the positions and the speeds of all particles through a position and speed updating formula, and limiting the particles exceeding the position and speed threshold after updating.

Step 5: and (3) iterating all particles, outputting the optimal fitness and the optimal different period information x1, x2 and x3 if the iteration times are completed, and otherwise, returning to the step (2).

The optimal heteroperiodic variables x1, x2 and x3 and the early warning time of the 4 data sets are finally determined through experiments, and the information is as follows:

data set	x1	x2	x3	Early warning time
					Misalignment 1	9	19	25	2802
Misalignment 2	2	3	21	1495
					Misalignment 3	7	16	14	1389
Imbalance 1	7	7	70	1923

The data distribution of the different period peak indexes can be seen in fig. 9, and the comparison of the different period peak indexes and the national standard peak monitoring effect can be seen in the following table:

as can be seen from fig. 8, 9 and the table above, the peak index of different periods can not only give an alarm in advance in the supervision process, but also reflect the fluctuation degree of each data segment, compared with the national standard peak-to-peak value real-time monitoring effect, the peak index has a larger improvement, and can leave sufficient time for the operation and maintenance of equipment.

2. Off-period monitoring for off-period pulse indicator

Rotor rub-impact fault data and shafting misalignment fault data are obtained, wherein a rotor fault part is a compressor, rated power is 674KW, rated rotating speed is 9508rpm, rated steam inlet pressure is 3.6Mpa (A), and rated steam exhaust pressure is 1.1Mpa (A). Unit anomaly description: abnormal fluctuation is frequent when the steam turbine operates, the amplitude is high, the amplitude exceeds an early warning threshold, friction is found at the shaft seal after maintenance, the steam turbine is restarted after cleaning, and the vibration value is stable; misalignment faults are described as: and a poor centering state exists between the low pressure cylinder and the high pressure cylinder.

According to the conventional method (national standard GBT 11348.1-1999 measurement and evaluation of radial vibration of rotating machinery rotating shaft), the vibration magnitude needs to be evaluated according to the peak value of displacement peak. And calculating a displacement peak value according to the extracted displacement data to obtain experimental data of the whole life cycle of each bearing. The data for all bearings are plotted as images with time represented by the abscissa and the ordinate amplitude refers to the peak-to-peak value as shown in fig. 10-15.

The data for the pre-warning using the conventional method are as follows:

fault data set	Early warning time/min	Total time/min of sample
			Rotor rub against 1	750	2000
Rotor rub against 2	700	2000
			Rotor rub against 3	950	2230
Misalignment 1	1225	1500
			Misalignment 2	1725	2000
Misalignment 3	1750	2000

According to different fault types under different working conditions, the data set is divided into a test set and a training set, and the data set is divided into the following tables:

in general, in normal operation of the device, the different period pulse index is relatively stable, generally, the different period pulse index does not exceed a certain ratio, the historical normal different period pulse index data is averaged once at intervals (for example, 20 min-100 min), the obtained average is averaged again, the baseline is taken as the baseline, and the baseline is increased by 15% to be taken as the initial early warning threshold (early warning threshold). See the fault dataset early warning thresholds shown in the following table:

data set	Rotor rub against 1	Rotor rub against 2	Rotor rub against 3	Misalignment 1	Misalignment 2	Misalignment 3
							Early warning value	1.15	1.5	1.31	3.1	3.0	2.3

The early warning time of each fault data set is obtained by adopting the embodiment, and compared with a national standard method, the early warning time is shown in the following table as the training test result of the fault data set:

as can be seen from training results in the table, the early warning time for fault monitoring based on the different-period pulse index is advanced by several tens of minutes or hundreds of minutes compared with the early warning time of the conventional method (national standard), and the period with the longest early warning time corresponding to each data sample in the table is extracted as the final period evaluation standard, and the period with the longest early warning time corresponding to the following table is extracted:

Data set	Rotor rub against 1	Rotor rub against 2	Rotor rub against 3	Misalignment 1	Misalignment 2	Misalignment 3
							Large period of time	50	30	50	50	70	70
Small cycle time	8	8	8	3	3	5
							Advance time/min	65	94	122	171	175	223

Experimental data show that the method for detecting the abnormal period provided by the invention can lead the early warning time to be tens to hundreds of minutes earlier than the national standard early warning time, and can make up for some losses caused by insufficient early warning time.

3. Aperiodic monitoring with respect to an aperiodic margin indicator

The method comprises the steps of obtaining a misalignment fault data set of a unit, wherein eddy current displacement sensors are adopted for measuring vibration of each unit, and the data collected by the eddy current displacement sensors are horizontal vibration or axial vibration displacement. And processing the displacement data by using a displacement peak-to-peak value to obtain peak-to-peak value period experimental data of each group of data, such as data of 8 measuring points shown in fig. 16-23, wherein the abscissa is time (unit: min) and the ordinate is displacement data based on the peak-to-peak value.

The alarm value (early warning threshold) may vary considerably from one machine to the next. The selected alarm value is typically set relative to a baseline amplitude that is determined empirically for the location and orientation of the station for the particular machine or machines of the same type. It is recommended to set the alarm value higher than the base line value by a value equal to 25% of the upper limit value of the area B. If the baseline value is low, the alarm value may be below region C. The alarm value can be set differently for different bearings on the machine, reflecting the difference in dynamic load and bearing support stiffness. In either case, the recommended alarm value should generally not exceed 1.25 times the upper limit of zone B. If the steady state baseline value changes, the alarm value settings need to be modified accordingly.

Therefore, in combination with the national standard and the actual situation of the data set, the situation that the data has deviation in actual operation is considered, a certain margin is required to be reserved for dynamic updating of the model, and the early warning value can be set to be 1.2 times of the baseline data, so that the corresponding monitoring early warning result is shown in the following table.

Dividing the data set into a test set and a training set, taking measuring points 1-3 and 5-7 as the training set, taking measuring points 4 and 8 as the test set, and training according to the data to obtain a proper different period value and an alarm value.

The machine is monitored through the sensor to obtain monitoring data, the monitoring data is processed according to an abnormal period margin index formula, baseline data of an abnormal period margin index is obtained in a stable operation stage of the equipment, and an early warning value can be set to be 1.2 times of the baseline value of the abnormal period margin index according to the baseline data. The early warning value is not a fixed value and can be adjusted along with the continuous updating of the data.

According to the corresponding size period, training results of all training sets are shown in the following table:

/>

as can be seen from the table, the early warning time after the abnormal period treatment is earlier than the untreated early warning time, and the period of the group with the largest early warning time in each group of data set in the table is extracted as shown in the following table:

According to the table, the early warning time represented by the large period 60 and the small period 2 is most advanced in all training sets, and the test set is tested according to the early warning time to verify the accuracy of the test set.

The verification process is as follows: the large period 60 and the small period 2 are tested on the test set, and the result is compared with the preprocessed data, and the obtained result is shown in the following table:

(Unit: min)	Early warning time of original data	Early warning time in verification process	Advance time	Total interval length
					Measuring point 4	440	318	122	1500
Measuring point 8	283	180	103	1000

From the above table, it can be seen that the large period 60 and the small period 2 have good early warning time advance effect on the test set, and early warning can be performed when the data has just occurred mutation, namely, in the fault symptom stage.

4. Aperiodic monitoring for aperiodic waveform indicators

The collision and grinding type data are selected for analysis, the obtained unit is a certain geothermal electric device, namely a No. 4 steam turbine generator unit and a steam turbine, and the rated power of basic parameters of the unit is 100MW; rated operating speed 3000rpm; and (3) unit coding: 140529044807131. a local circulating hydrogen compressor (H2275) -a steam turbine, wherein the basic parameter rated power of the unit is 674KW; rated operating speed 9508rpm; and (3) unit coding: 170318044053924. the system sampling interval is once every five minutes, vibration displacement of the rotating shaft is acquired through a displacement sensor, and the acquired relevant time, frequency, speed, waveform data and the like at the current moment are stored in a txt file. The fault part of the data set 140529044807131 in the collision and grinding mode is that vibration fluctuation exists at the high-pressure non-connection end of the steam turbine, and the system report is 125; the fault part of the unit 170318044053924 in collision and grinding is that the high-pressure non-linkage end of the steam turbine has vibration fluctuation, and the system height report is 66.5. Wherein we choose the set 140529044807131 as the training set and the set 170318044053924 as the test set.

The early warning setting is carried out according to the actual situation and the boundary of the national standard evaluation area, and the appendix of the national standard GBT 11348.2-2012 mechanical vibration measuring and evaluating machine vibration 2 nd part on a rotating shaft gives that the boundary of the evaluation area of the steam turbine and the generator which are installed on land with rated rotating speeds of 1500rmin, 1800rmin, 3000rmin and 3600rmin is about more than 50MW, and is shown in figure 24.

And preprocessing error data of the original data set, and carrying out time domain dimensional index displacement peak-peak analysis. The data collected by the dataset is stored in dictionary form in each txt file, collected every five minutes. Through traversing txt files at each moment, waveform data in extraction are extracted to obtain peak-peak values, and early warning monitoring of displacement peak-peak values of the training set test set time domain dimensional indexes under national standards can be drawn, such as displacement peak-peak value curves shown in fig. 25 and 26, wherein the training set 140529044807131 peak-peak early warning time T0=2855 and the testing set 170318044053924 peak-peak early warning time T0=705.

And calculating training set data of the original data set according to a calculation formula of the different-period waveform index. Firstly, dividing the size period of the different period, and dividing the small period into 20, 40 and 80 by the original data set; the large period is 100, 200, 300.

The dimensionless waveform index is established through different periodicity to serve as a monitoring index, the average value of the waveform index is 1 under the condition of normal operation of the unit, and the waveform index is set to be an early warning value (early warning threshold value) when the index exceeds 30% of the average value of the waveform index, so that the early warning value G1=1.3 of the different periodic waveform index can be set. According to the early warning monitoring of the actual situation under the national standard of the displacement peak-peak value of the dimension index, the peak-peak value early warning time T0=2855 can be obtained, the abnormal periodic waveform index early warning time is T1, and the advance time H=T0-T1. Since the dataset is measured every five minutes, the actual advance time h1=hx 5. The results of the different period training are shown in the following table:

as can be seen from the selection of different size periods in table 1 for monitoring and early warning, when the small period is unchanged and the large period is continuously increased, the monitoring and early warning time of the system is increased and then reduced, and the early warning time is reduced and then increased, such as the small period 40, the large period 100, 200 and 300. The early warning time effect is better when 300 early warning time is selected in a large period. When the large period is unchanged, the monitoring and early warning time of the system is continuously increased along with the continuous increase of the small period, and the early warning time is continuously reduced, such as the small periods 20, 40 and 80 and the large period 300. The early warning effect is better when 20 early warning time is selected in a small period. By comparing the early warning time, the small period 20 and the large period 300 are selected as new monitoring early warning index parameters.

The unit 170318044053924 is used as a test to verify the small period 20 and the large period 300, the early warning time is 705 under the early warning index of the time domain dimensional peak and peak value of the original data set, and the different period test result obtained under the new waveform index through the verification set is shown in the following table:

through verification, the original data set has the early warning monitoring of the dimensionality index displacement peak-peak value under the national standard, the abnormal periodic waveform index is obtained through carrying out abnormal periodic dimensionless construction on the dimensionality index, the monitoring early warning time is effectively advanced compared with the original monitoring early warning time, and the feasibility of the abnormal periodic waveform index monitoring early warning method is verified through experimental training test.

5. Aperiodic monitoring of an aperiodic kurtosis index

The data adopted by the invention is provided by Shenyang blower group measurement and control technology Co., ltd, and the fault type is shafting misalignment fault. The sampled data type is real-time vibration displacement data of the unit, the sampling interval is about 35min, the sampling time is 1s, and the sampling frequency is 1024Hz. The data of a whole part is divided into two parts of a training set and a testing set. While partial training set data represents the machine misalignment case as follows:

And the test set data fault conditions representing sensor faults are shown in the following table:

each sample had 1024 sampling points, and the original data of the 4-group misalignment fault is shown in fig. 27:

for the data, different machines are set according to national standard early warning value setting standards, dimensional characteristics in the machines are extracted for better comparison, peak-to-peak values are selected as characteristic quantities for collection, and the peak-to-peak value characteristics and peak-to-peak value monitoring early warning time of each machine with non-centering faults are shown in the following table in fig. 28:

according to the calculation formula of the different period kurtosis value ratio, calculating the data of the training set, and early warning the monitoring index (different period kurtosis value) H ₁ Setting; in addition, the situation that accidental impact or artificial interference can exist in the actual production process of the rolling bearing is considered, and the early warning value H can be obtained ₁ Further increase and set upSetting the early warning value of the second monitoring index as H ₂ (H ₂ >H ₁ ). When the value of the monitoring index exceeds the early warning value H ₁ Or H ₂ And (3) early warning is carried out on the device.

For example, assuming that x1' =30, x2' =40, x3' =180, an early warning is performed on the non-centered data set, and a first monitoring index early warning value H is set ₁ On the basis of the value of 1.1, the H value can be further increased by considering the accidental impact or artificial interference, and the early warning value of the second monitoring index is set as H ₂ =1.2. When the value of the monitoring index exceeds the early warning value H ₁ Or H ₂ And (3) early warning is carried out on the device.

Different dynamic parameters (H, x ', x2', x3 ') lead to early warning time advance or late, so that optimization of the monitored index parameters is required. Optimizing the monitoring index parameters by a method of repeated cycle comparison, comparing the monitoring index parameters with the early warning time of the peak value, training in a training set, and taking the optimized group of monitoring index parameters as the monitoring index parameters of a subsequent test set. And carrying out secondary correction on the dynamic parameters, so that the highest speed early warning is achieved on the premise of no false warning.

The results obtained for the dataset through multiple training are as follows:

the results of the tests on all training sets are shown in the following table (bolded to indicate best results):

from this, it is known that the warning time of the dimensionless monitoring index tends to increase or remain unchanged with the increase of the short period x 1'. In the case of multiple tests, the early warning caused by too small early warning line or x1' arrangement can also occur, so that false warning is generated Thus, early warning is difficult to be effectively performed in advance. Results that are more stable and effective than dimensional monitoring are not achieved. Because there is an offset of x2 '(x 2'>x1 ') so that signals of short period and long period are not overlapped, but along with the continuous increase of long period parameter x3', dimensionless early warning and monitoring time is continuously delayed. The value of the long period x3' should not be too large either. Because the dimensionless index prediction needs to be applied to actual production, accidental impact caused by the working condition of the actual production and a part of artificial operation errors needs to be considered. This requires an up-regulation of the early warning value H, which better reduces false alarms. Two sets of monitoring index parameters are thus selected here: the first group is: h ₁ =1.1, x1' =30, x2' =40, x3' =120. The second group is: h ₁ ＝1.2,x1’＝30,x2’＝40,x3’＝180。

The test set is then tested: calculating the divided test set, and obtaining a structure shown in the following table according to the recommended monitoring index and the different period calculation formula:

test set number	H ₀ Early warning line	Early warning time
			15	1.2	151
17	1.2	1444
			19	1.2	151
20	1.2	528

Test set number	H ₀ Early warning line	Early warning time
			15	1.1	1472
17	1.1	845
			19	1.1	101
20	1.1	577

The first group of monitoring index parameters can be obtained from the test results, and the monitoring of the rolling bearing is already met, and the advance time is better; thus, select A first set of monitoring index parameters, namely: h ₁ ＝1.1、x1’＝30、x2’＝40、x3’＝120。

In one embodiment, the data processing end is further configured to apply the abnormal period monitoring model to online monitoring of the mechanical device, and update the model periodically according to an online monitoring result.

In one embodiment, the data processing end is further configured to update the model by:

In one embodiment, the model is updated by:

and optimizing the abnormal period monitoring model by taking false alarm data and missing alarm data which are monitored on line as training data to obtain an updated abnormal period monitoring model.

It can be understood that real-time online dynamic monitoring and early warning are achieved in the continuous running process of the unit, meanwhile, the situation that false alarm and missing alarm occur in the fault abnormality of the unit is effectively prevented, and the protection function is achieved. And when the false alarm condition occurs, the early warning threshold is reduced, and when the false alarm condition occurs, the early warning value is increased.

In one embodiment, the data acquisition end comprises a speed sensor, a signal conditioning module, a signal filtering module and an acquisition card;

In one embodiment, the data acquisition end is a three-axis acceleration sensor or a wireless vibration sensor.

For different rotary machines, the fault monitoring is performed by adopting different forms of systems, including a PC end time-frequency domain fusion intelligent online fault diagnosis system, a handheld vibration detector and a wireless vibration sensor vibration monitoring system.

Scheme one: and the PC end time-frequency domain fusion intelligent online fault diagnosis system. Referring to fig. 29, the scheme adopts a time-frequency domain fusion intelligent online fault diagnosis system formed by a collection device and PC end software, and is mainly used for diagnosing large-scale rotating machinery, such as a turbine unit, a flue gas turbine unit, a rubber device extrusion dehydrator and an expansion dryer. The system collects and processes vibration signals through a collecting device (a data collecting end), and a server (a data processing end) stores and manages data and uploads and issues the data on the network; the collected online actual measurement data of the industrial field large-scale equipment of the enterprise user is transmitted to a remote fault diagnosis center through a remote network, then the data is processed by utilizing a rotary mechanical compound fault intelligent diagnosis algorithm, and further the working state, the development trend and the characteristics of compound faults of the industrial field equipment are analyzed, so that the experimental verification and the application research of the state trend prediction and the compound fault diagnosis of the industrial field equipment are realized. The acquisition device mainly comprises a speed sensor, a signal conditioning module, a signal filtering module and an acquisition card. The speed sensor is simple to install, is insensitive to the installation surface, is used for medium-frequency measurement, is suitable for vibration measurement and monitoring of rotary machinery, and can effectively detect rotor unbalance, oil film vibration, rolling bearing faults and the like. However, the life is limited and the springs (membranes) suspending the magnets inside the speed sensor wear. Due to the influence of the spring resonance, it has an upper frequency limit and a lower resonance frequency. Mounting position: is installed in the horizontal and vertical directions of the rotary machine bearing cap. In order to obtain real and effective vibration data of the system, the acquisition frequency is set to be 8kHz. The data acquisition process comprises the following steps: the speed sensor converts the physical quantity generated by the vibration of the rotary machine into a weak electric signal, the weak electric signal is transmitted to the signal conditioning module, the signal conditioning module amplifies the received weak signal, the amplified signal is provided for the signal filtering module to filter out the high-frequency signal, and then the signal is transmitted to the acquisition card, and the acquisition card acquires and processes data and uploads the data to the PC end for processing.

Scheme II: a hand-held vibration analyzer. Referring to fig. 30, the scheme adopts an acceleration sensor and a touch screen (data processing end) to form a handheld vibration analyzer. The acquisition device mainly comprises an intelligent triaxial acceleration sensor, a direct-current power supply adapter (used for converting AC220V into direct-current voltage to supply power for a touch screen and the acceleration sensor), an RS485 communication module (used for communication between the triaxial acceleration sensor and the touch screen) and the touch screen. The system is mainly used for monitoring small and medium-sized rotary machines, such as small and medium-sized fans, small and medium-sized motors, small and medium-sized pumps, small and medium-sized air compressors and the like, and can rapidly diagnose faults of equipment. The triaxial acceleration sensor is simple to install, can detect wide-range frequency vibration, comprises short-time impact, has good effects on detecting rotor imbalance, rotor misalignment, shaft bending, lost or damaged parts, component resonance and the like, and is an excellent sensor for detecting faults of rolling bearings and gears. Its service life is also long. Mounting position: the mounting surface of the sensor and the measured surface are fixed tightly, smoothly and stably; the axes of the sensor and the measured axes must be parallel, and the two axes cannot form an included angle as far as possible. The handheld vibration analyzer is mainly used for rapidly detecting faults of the rotary machine on line, so that the acquisition frequency is 2kHz, and the online fault diagnosis function can be met. The data acquisition process comprises the following steps: the touch screen is used as an upper computer, the intelligent triaxial acceleration sensor comprises a data processor as a lower computer, the sensor converts physical quantity generated by rotary mechanical vibration into an electric signal, then the electric signal is converted into an analog signal through processing of an internal processor, the analog signal is sent to the upper computer of the touch screen through an RS485 module, an online vibration monitoring system (a data processing end) in the upper computer can conduct online analysis on the sent vibration signal, and a fault result can be diagnosed rapidly.

Scheme III: a wireless vibration sensor vibration monitoring system. Referring to fig. 31, the wireless vibration sensor vibration monitoring system is formed by adopting a wireless vibration sensor, a server and a time-frequency domain fusion intelligent online fault diagnosis system. The acquisition device mainly comprises a wireless vibration sensor, a communication transponder and a power supply (for supplying direct-current voltage to the wireless vibration sensor). The system is applicable to vibration monitoring of most rotating machinery, and vibration data of the wireless vibration sensor can be uploaded to a server through a communication transponder, so that the intelligent diagnosis algorithm of the composite fault of the rotating machinery can be utilized for data processing, the characteristics of the working state, the development trend and the composite fault of the industrial field equipment are further analyzed, and experimental verification and application research of the state trend prediction and the composite fault diagnosis of the industrial field equipment are realized. The wireless vibration sensor is internally provided with an acceleration sensor module, has a larger vibration frequency detection range, has good effects on detecting rotor imbalance, rotor misalignment, shaft bending, lost or damaged parts, component resonance and the like, and is an excellent sensor for detecting rolling bearings and gear faults. The wireless vibration sensor mainly comprises an acceleration sensor module, a data processing module, a wireless transmission module, a display screen and a battery. Vibration data can be displayed on a display screen of the sensor in real time, and on-site overhaul and maintenance of equipment are facilitated. Wireless vibration sensor mounting position: the horizontal and vertical directions of the installation on the rotary machine bearing cover or the shell are consistent with the installation position of the speed sensor. In order to enable the intelligent online fault diagnosis system to effectively diagnose faults of the rotary machine, the acquisition frequency is set to be 8kHz. The data acquisition process comprises the following steps: the wireless vibration sensor converts the physical quantity generated by the vibration of the rotary machine into an electric signal, the built-in data processing module of the sensor converts the electric signal into a digital signal, the display screen on the sensor can display the acquired vibration data in real time, the digital signal is transmitted to the communication transponder through the wireless transmission module inside the sensor, the communication transponder amplifies and transmits the signal to the server, and finally the data is analyzed and processed through the intelligent online fault diagnosis system (data processing end) and fault diagnosis is carried out on the machine.

Compared with the prior art, the mechanical equipment fault monitoring system provided by the embodiment of the invention obtains the non-dimensional index of the different periods by selecting the different dimensional parameters of the different periods to be used for analyzing the mechanical equipment fault, effectively inhibits the interference of the external environment, and simultaneously improves the defect of untimely early warning.

While the foregoing is directed to the preferred embodiments of the present invention, it will be appreciated by those skilled in the art that changes and modifications may be made without departing from the principles of the invention, such changes and modifications are also intended to be within the scope of the invention.

Claims

1. A mechanical equipment fault monitoring system, comprising:

2. The mechanical equipment fault monitoring system of claim 1, wherein the off-cycle dimensionless number is an off-cycle peak number, the first dimensionality parameter is a signal peak value, and the second dimensionality parameter is a root mean square value; the off-period peak indicator is equal to a signal peak of the vibration data at the first period divided by a root mean square value of the vibration data at a corresponding second period.

3. The mechanical equipment fault monitoring system of claim 1, wherein the out-of-period dimensionless number is an out-of-period pulse number, the first dimensionality parameter is a signal peak, and the second dimensionality parameter is an absolute average; the off-period pulse index is equal to the signal peak value of the vibration data in the first period divided by the absolute average value of the vibration data in the corresponding second period.

4. The mechanical equipment fault monitoring system of claim 1, wherein the aperiodic dimensionless index is an aperiodic margin index, the first dimensional parameter is a signal peak, and the second dimensional parameter is a mean square value; the abnormal period margin index is equal to a signal peak value of the vibration data in the first period divided by a mean square value of the vibration data in a corresponding second period.

5. The mechanical equipment fault monitoring system of claim 1, wherein the aperiodic dimensionless index is an aperiodic waveform index, the first dimensional parameter is a root mean square value, and the second dimensional parameter is an absolute average; the abnormal-period waveform index is equal to a root mean square value of the vibration data in the first period divided by an absolute average value of the vibration data in a corresponding second period.

6. The mechanical equipment fault monitoring system of claim 1, wherein the off-cycle dimensionless index is an off-cycle kurtosis value ratio index, the off-cycle waveform index being equal to an average of kurtosis values of the vibration signal at the first cycle divided by an average of kurtosis values of the corresponding second cycle.

7. The system for monitoring a fault in a mechanical device according to any one of claims 1 to 6, wherein the data processing end is further configured to obtain the first period, the second period, and the early warning threshold by:

constructing an original abnormal period monitoring model;

8. The system for monitoring mechanical equipment failure according to claim 7, wherein the data processing terminal is further configured to apply the abnormal period monitoring model to online monitoring of the mechanical equipment, and periodically update the model according to an online monitoring result.

9. The machine fault monitoring system of claim 8, wherein the data processing end is further configured to:

10. The mechanical equipment fault monitoring system of claim 1, wherein the data acquisition end comprises a speed sensor, a signal conditioning module, a signal filtering module and an acquisition card;

11. The mechanical equipment fault monitoring system of claim 1, wherein the data acquisition end is a three-axis acceleration sensor or a wireless vibration sensor.